Why did AMD surge by 9% overnight, with a market value exceeding 300 billion?

On February 29th, Thursday, AMD's stock price surged over 9%, pushing its market value to surpass the $300 billion mark for the first time, reaching $311 billion. In after-hours trading, AMD's stock further rose by 1.93%.

Driven by AMD's stock surge, chip stocks in the US also rose, with the PHLX Semiconductor Index closing up by 2.7%. NVIDIA, as a "substitute stock," performed well, with stocks like Super Micro Computer Inc. (SMCI) and Marvell Technology Group Ltd. (MRVL) rising by 6.1%, and ARM's stock increasing by 5.4%.

Earlier on Thursday, Citigroup stated that with the continuous development of AI technology and the expanding application areas, the demand for AI chips from various enterprises and organizations is increasing. Citigroup remains "extremely bullish on the semiconductor industry" and has listed AMD, NVIDIA, and Broadcom as some of its favored stocks.

Furthermore, Dell Technologies made optimistic comments about GPU demand on Thursday afternoon, which continued to boost AMD's stock price in after-hours trading. Dell Technologies' Chief Operating Officer Jeff Clarke stated during the earnings conference call:

"We are seeing strong interest and orders for AI servers equipped with AMD and NVIDIA's next-generation AI GPUs, including NVIDIA H200 and AMD's MI300X."

Due to the surge in demand for AI servers, Dell Technologies achieved strong quarterly performance, driving its stock price up by nearly 15%. As AMD is one of the GPU suppliers for Dell Technologies' servers, Dell Technologies' statement indirectly increased market confidence in AMD's products.

Strong Growth in AMD's Data Center GPU Business

AMD is considered one of the major beneficiaries of artificial intelligence. Data shows that in the past few years, AMD's ranking in the S&P 500 Index has significantly risen from 222nd five years ago, 114th four years ago, 58th one year ago, to the current 22nd.

On Tuesday, AMD released its fourth-quarter and 2023 financial reports. The Q4 revenue was $6.168 billion, a 10% increase YoY; the full-year 2023 revenue was $22.68 billion, a 4% decrease YoY. When looking at it by department, the revenue of the Data Center Business Department reached $2.282 billion, a year-on-year increase of 38%. The main reason for the revenue growth of the Data Center Business Department is the increased shipments of the fourth-generation EPYC processors for data centers and the Instinct GPUs for AI and high-performance computing applications by AMDD.

Lisa Su, the CEO of AMDD, stated that the growth in the data center GPU business due to the increased demand brought by the artificial intelligence boom led to impressive financial performance in Q4:

"Thanks to the significant increase in sales of AMDD Instinct GPUs, EPYC CPUs, and Ryzen processors, the company achieved both MoM and YoY revenue and profit growth in 2023."

"Moreover, with the continued acceleration in demand for AMDD's high-performance data center product portfolio, AMDD is in a favorable position to leverage the development and application of artificial intelligence technology to further drive the company's growth."

However, AMDD's performance guidance fell below expectations, causing its stock price to drop by 6% in after-hours trading following the earnings report. AMDD's performance guidance indicates that although the shipments of the client platform and gaming hardware are expected to decline, the stability of its data center business and the anticipated growth in AI and HPC GPU sales are expected to help maintain overall revenue levels, as AMDD gradually shifts towards more growth-oriented market areas.

Given the shortage of NVIDIA AI GPUs, many expect AMDD to face similar issues when launching the MI300 series AI GPU. However, AMDD has stated that it will not be constrained by supply limitations and is capable of meeting market demand for the MI300 series AI GPU. AMDD expects to sell MI300 series AI GPUs worth $3.5 billion by 2024.

AMDD is committed to strengthening its leading position in the inference market

On February 28, during a meeting with Bernstein, AMDD's Chief Technology Officer emphasized the company's strategic positioning in the AI and HPC markets, as well as its advantages in technology, products, pricing, and ecosystem. AMDD is dedicated to enhancing its leading position in the inference market, while meeting customer demands through accelerated product updates and optimized software support:

1. Cloud Training and Inference: Most AI training tasks are conducted in the cloud, and the scale of inference (the actual application of AI models) will be much larger than training. AMDD holds a leading position in the inference market, with larger memory capacity and a first-mover advantage compared to NVIDIA. The inference market includes data centers (DC), enterprises, and edge/embedded fields, where AMDD has a better ecosystem and market position. 2. Pricing Strategy: The gross profit margin (GPM) of AMDD's 300x GPU is similar to its CPU products. Even in the face of NVIDIA's new product releases, AMDD will not change its pricing strategy, but will focus on the total cost of ownership (TCO).

3. Software Support: Large Language Models (LLM) are lowering the barrier to software development. Software companies are changing their coding methods, and AMDD's compiler provides good support for this.

4. Depreciation Cycle of AI Servers: AI servers are expected to have a longer depreciation cycle, and power efficiency will be a key factor in deciding whether to replace servers.

5. National-Level AI/HPC Applications: High-Performance Computing (HPC) and AI are becoming more and more similar, with more AI functionalities being integrated into HPC. More and more national laboratories are using AI, bringing good development prospects to the market, where AMDD has an advantage.

6. ASIC Development: AMDD is rebuilding its custom semiconductor team, shifting from the gaming field to the data center and server domains.

7. Product Release Cadence: Driven by customers, AMDD is accelerating the pace of product releases. Although customer testing and adoption may take up to 18 months, the release cycle will not be shortened to 12 months.

8. Cooling and Power Supply: With technological advancements, cooling and power supply are becoming increasingly important. AMDD currently uses industry-standard solutions provided by OEMs.